Report #9548
[agent\_craft] Agents repeat failed tool calls without analyzing error messages, causing infinite loops
Implement structured error reflection: when a tool returns an error, force a reflection step that categorizes the error \(transient vs permanent, syntax vs logic\), quotes the specific error substring, and generates a revised plan before retrying; limit to 3 attempts with exponential backoff
Journey Context:
Simple retry loops fail because they don't condition on error content; Reflexion research shows that agents with explicit self-reflection steps recover from errors 30% more often. The key is forcing the model to 'stop and think' rather than immediately generating the next tool call. This prevents the 'action loop' failure mode where agents spam the same broken command.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T08:24:32.934543+00:00— report_created — created